Google NotebookLM: Creating Engaging Language Lessons
By Priya Panday-Shukla, Washington State University
DOI: https://www.doi.org/10.69732/DZWQ2675
Intro
I tried Google NotebookLM about a year ago, and I thought it was just another AI tool; now, almost a year later, I have to say that it has emerged as something unexpectedly brilliant for education, especially for language learning, for the way it transforms everyday content into engaging lessons. Teachers can upload their preferred topic in the form of web links, texts, or audio, and the tool will convert them to the content of choice (e.g., summary, study guide) and even create natural podcast-style conversations.
Overview
NotebookLM is an experimental AI-powered tool developed by Google Labs. You can upload different types of materials and interact with them in a conversational manner through a chat interface and create a wide range of content, including a podcast-style audio overview. NotebookLM runs under Gemini 1.5 Pro, Google’s most powerful model as of July 2024. It is possible to upload up to 50 sources with a 500,000 words per source limit (at the time of writing this article). Here is what makes NotebookLM so special: it only works with the materials you give it; it utilizes retrieval augmented generation (RAG) techniques, meaning that it pulls information from the uploaded sources, which creates less chances of hallucinations (i.e., making things up). It cites which source it is taking the information from, making the information quite reliable compared to other language models.
In its privacy and security notice, NotebookLM states that it will not use the uploaded data to train its model. If you send feedback, human reviewers might check the information to address any concerns, and therefore, it is recommended to not include any personal or confidential information within the sources or chat.
Name of tool | Google NotebookLM |
URL | https://notebooklm.google.com/ |
Primary use of tool | Synthesizing large amounts of data from diverse sources |
Cost | Free; it is in an experimental phase |
Ease of use | Easy to use |
How to Access and Use the Tool
Currently, NotebookLM is available freely to all users over 18 in the regions where Gemini API is available. You can access the tool on a web browser as well as on mobile devices, but the desktop experience is recommended for better performance. You can start by visiting https://notebooklm.google.com/, sign in with their Google email, and begin by creating a “New Notebook” and importing your desired sources. You have the ability to incorporate up to 50 sources, including Google Docs and Slides, PDFs, websites and YouTube links, texts, and audio files.
Once the sources are uploaded, the tool offers a brief summary of the uploaded materials, and clicking on the “Notebook Guide” button provides options to generate frequently asked questions (FAQs), study guides, a table of contents, timelines, briefing docs, and even an audio overview that sounds like a very realistic podcast conversation (which is, actually, two AI voices speaking in English). For the text and chat, NotebookLM will answer in your default language, so make sure to add a second or third language that you are teaching or learning in your Google account settings. Then, ask the tool to translate, for example, a summary into your preferred language.
The audio overview is by far one of the most exciting features since it can be used to ignite engagement in learning. Generating the podcast takes a few minutes, but it is worth waiting for; the resulting audio is as good as (or dare I say, even better than some) real podcasts. Currently, it only supports English, but Google recently added a “Customize” button where you can input some instructions to create the podcast in other languages. To try it, I uploaded only four sources about tostones (three webpages and a video) and added the following instructions: “This podcast is for students learning Spanish as a second language. We need the hosts to speak in Spanish from Latin America since tostones are from this continent, and the podcast should sound authentic. The hosts should focus on how to make tostones. Use simple and easy steps so students can follow along,” and it gave me a 4:29 audio clip that sounds like authentic Spanish (listen to the audio).
I can see this podcast being used in the Spanish language classroom for students to listen to a specific regional accent while following this traditional recipe. You can click on the vertical ellipsis to adjust the playback speed if needed; you can also interact with the sources by asking questions in the chat box, and the AI will process the request. For example, try asking for summaries of specific information, comparing information from diverse sources, or looking for trends within the materials. The tool will pull the information from within the uploaded sources into that specific notebook; and if it does not have the answer, it will state it and will not make up the answer to please you. As of right now, the podcast can be generated only once (no edits). The only way to get the transcript would be to download the audio file and then upload it (when audio and video links are uploaded the system generates a transcript). While NotebookLM can be a powerful teaching, learning, and research tool, its generated content, as with any AI-created content, needs to be fact-checked and edited before sharing it with others.
Suggested Uses for Language Teaching and Learning
NotebookLM can be used in different aspects of language education. Below you can find several examples.
- Content creation: Teachers can upload news articles, blog posts, video links, and audio and ask the tool to synthesize the information or outlines for a presentation. This can be used at the beginning of the class, or in a flipped classroom, for students to learn quickly about the topic and listen to the expressions of diverse accents.
- Summaries: Teachers can upload articles and video links about festivities and how they are celebrated in different countries, for example. Ask the tool to summarize the information based on the cultural significance and ways it is celebrated today. Versions with differing proficiency levels can be easily created.
- Collaborative learning: Students can share notebooks and collaborate, for example, on projects exploring different aspects of the target language cultures by investigating historic contexts and contemporary media, and compiling the findings in the notebook. Then, the information can be synthesized and presented to the class.
- Podcast: Teachers can upload text, audio, or video links based on a particular topic or theme and ask the tool to create a podcast for them. For example, teachers could use travel blog websites, video links, and even audio to create an engaging podcast in the target language by clicking on the “Notebook Guide” feature, then “Customize” (here is where you tell the tool that it needs to generate the podcast in a specific language), and finally, “Generate” (located in the Deep Dive Conversation box).
- Research: Teachers can upload technology news websites in the target language, for example, websites and tech forums in Spanish and articles that talk about Spanish language evolution. Ask NotebookLM to generate a list of new tech vocabulary in Spanish, regional variations, and new Spanish terms for digital concepts (e.g., smartphone, móvil, celular).
- Study guide: Teachers and students can make use of this feature by uploading relevant materials like grammar points, vocabulary lists, classroom notes, and any supplementary articles related to the class. Then, click the “Notebook Guide” button followed by the “Study Guide” feature. This will create a guide with key terms and potential exam questions. Students can use the “Chat” interface to ask the tool specific questions depending on their learning needs.
- Discussions: Teachers can upload text or video links on current events or cultural topics and ask NotebookLM to create engaging discussion questions to promote critical thinking and allow students to express themselves in the target language or their views of the language they are learning.
Advantages
- Generates summaries, study guides, and engaging audio overviews that can be used for language teaching and learning.
- Saves time to create engaging materials for the world language classroom with quick customization for learning objectives and proficiency levels.
- It pulls information from the uploaded sources and cites which source the generated text it is taking it from, making it easy to track and verify information.
- It can process multiple sources and formats efficiently (e.g., PDFs, audio, web links).
- Free of cost to all users; making learning more equitable for everyone.
Limitations
- It is important to note that the chat history does not save between sessions, but it can manually be saved as a note.
- Asking it to translate the English text into Spanish does not always work smoothly; sometimes you might have to try two or three times, though it always worked correctly for the audio podcast.
- Although it pulls the content directly from the sources, there is still potential for hallucinations, especially when it comes to the podcast since it fluffs the content to make it more engaging to its listeners. However, in my experience, the fillers it creates for the audio are not actually wrong information, it does follow the theme and add explanations that are connected with the sources. So far, I have not seen any hallucinations that it pulls from the sources, and as mentioned before, NotebookLM tells you if it does not have enough information to answer your queries.
Conclusion
NotebookLM offers diverse applications that can help enhance both teaching and learning experiences. From creating dynamic content and personalized study guides to exciting podcasts. It certainly has a lot to offer, especially its ability to generate on-demand content with specific regional accents, which can alleviate an important accessibility problem in language education. However, as with any technology, this AI tool should be used thoughtfully to really support pedagogical goals, augment learning, and foster engagement and language acquisition.
References
McHugh-Johnson, M. (2024, October 18). 8 expert tips to getting started with NotebookLM. Google. https://blog.google/technology/ai/notebooklm-beginner-tips/
Wang, B. (2024, September 11). NotebookLM now lets you listen to a conversation about your sources. Google. https://blog.google/technology/ai/notebooklm-audio-overviews/
I was very impressed with the quality of the Spanish podcast you posted. I had heard that it was possible to use the Customize the podcast specifying that it should be between two people speaking another language.
I am more interested in using a source created by a language learner; some writing or an audio recording or the transcript of one and then getting the learner to listen to the podcast, which is going to be about what they have said themselves, so it will be more intelligible than the level of the podcast, which is usually B2. This means it works well with B1 learners, especially if they can follow the transcript as they listen.
Personally, I upload the audio of the podcast to Turboscribe.ai or Rev.com and then get the student to listen to the podcast as they follow a synchronized transcript. The idea you suggest of downloading the audio of the podcast and uploading it to the same NotebookLM means that students can listen and follow the transcript on a PC. It IS possible on a mobile phone if you start the podcast first (and adjust the speed down to 80%) and then go to the sources and choose the source with the transcript.
I’ve written about this on my blog but here’s a link to one on NotebookLM
https://notebooklm.google.com/notebook/00880ff9-47e7-4f6e-8b51-403d32f209f8